[Mageia-dev] [changelog] [RPM] cauldron core/release kernel-3.7.0-0.rc8.1.mga3

Sun Dec 9 15:54:12 CET 2012

On 9 December 2012 15:00, Olivier Blin <mageia at blino.org> wrote:
+>>> - add perf bash_completion
+>>> - more filelist updates
+>>> - add 3.7 buildfixes for alx, IFWLOG, mach64, ndiswrapper
+>>> - pull in more upstream git fixes
+>>> - rediff disable-mrproper patch
+>>> - update to current rc8+ git
+>>> - update filelists
+>>> - update defconfigs
+>>> - restore patch preferring ata over ide drivers
+>>
+>> BTW kernel-linus is still affected.
+>> Of course this conflicts with the policy patch...
+>
+> Maybe something worth being submitted upstream then?
+
+Or changing aliases order in ldetect.
+With modules-init-tools, we used to pick the last alias:
+
+        // take the last one because we find eg: generic/ata_generic/sata_sil
+        struct module_alias *it = aliases;
+        while (it->next)
+            it = it->next;
+        result = strdup(it->module);
+
+with kmod, we pick the first one (b/c that's the order that kept the same
+results as with modules-init-tools, when I switched ldetect from unsupported
+libm-i-t to kmod):
+
+        kmod_list_foreach(l, list) {
+                struct kmod_module *mod = kmod_module_get_module(l);
+                //if (str) // keep last one
+                //      free(str);
+                if (!str) // keep first one
+                str = strdup(kmod_module_get_name(mod));
+        }
+
+It used to works nicely but it appears that it works only good with a
+patched kernel.
+When we drop our patches, letect broke, reporting the wrong results
+(https://bugs.mageia.org/show_bug.cgi?id=8315)
+
+We could drop that patch for good and pick the last found alias instead of
+the first one in ldetect.
+See:
+- the patch: https://bugs.mageia.org/attachment.cgi?id=3204
+- the result in lscpidrake output on unpatched kernel:
+  https://bugs.mageia.org/attachment.cgi?id=3205
+
+In fact, I'm considering submiting a patched ldetect with the changed ordering
+to core/updates_testing and ask people to report the difference (if any)
+between:
+- core/release's ldetect + kernel-desktop
+- core/updates_testing's ldetect + kernel-linus
+
+WDYT?
+